As I work for British telecom I have the knowledge how a call routes across the network from one yogurt carton along a piece of string to the other...if only if was that simple. Next time you report a fault to your telecom provider reporting noise when calling your Uncle Bill please be patient for it to be resolved as the routining across the network is not as simple as some people may think.
Well here is some information that I produced for a training session a couple of years ago, perhaps interesting for the nerds and useful for the insomniac....my mandatory techie part of my blog in the attempt to make myself sound more clever than what I really am :-)
A (regular) customer’s line runs underground or overhead on poles to the local BT building. While this building is often known as "the exchange", in actual fact it might well not be. All the lines in an area terminate on a large patchboard known as the Main Distribution Frame (MDF). A second set of wires run from the MDF to the switching unit in the building. For the majority of lines this unit is a Remote Concentrator Unit (RCU), while some lines terminate on a co-located concentrator (usually in the same building as the DLE, this is the same for a AXE10 as they use an RSS and a SSS for co-located)
Usual maximum size of a CCR/RSS is 2048 customers, served by Mk1 or 2 CCR which has max of 8 DPH links, to a MK3 which has a max of 16 DPH links. (same for RSS). The big Mk3 sites tend to deal with ISDN customers. There is a new developed CCR which has up to 32 DPH links, but these are quite rare. The RCU can connect calls between two subscribers, or it can connect a subscriber line to a DLE link. There are 6941 RCUs of three different types: 4339 Marconi System Xs, 2137 Ericsson AXE10s, and 465 UXD5s. Most RCUs are located in villages or in small towns, but a number are in the same building as the DLE that controls
There are 770 DLEs around the UK, consisting of 559 System Xs and 211 AXE10s (note that System Xs and AXE10s perform a number of different roles in the network). Each DLE has its own subscribers and also controls one or more RCUs, for a total of 30 to 40 thousand lines. The DLE controls all calls made between those subscribers, as well as taking part in calls to and from elsewhere. DLEs are not spread evenly across the country - indeed there are 149 in London but only 21 in the whole of Northern Ireland. Rather, they are distributed to match the density of lines. A building in a major town may well have two or three DLEs, each with its own cluster of RCUs. These will not necessarily divide the locality geographically - it's possible for adjacent villages to be on different DLEs while those far apart may be on the same one. Similarly, there is not necessarily a connection between DLE and dialling code: all four of the DLEs in Cambridge have both 01223 and 01954 numbers on them, and at least one has 01353, 01440, 01553, 01638, and 01842 numbers as well
A System X switch is a unit that has a MDF and trunks to other switches. Therefore an RCU is a switch that only connects to its DLE, while a DLE is a switch that connects to many other switches, including RCUs
With the AXE10, on the other hand, the MDF is connected to a separate switching unit called the SSS. The SSS can be either physically attached to the switch (in which case it is called a CSS) or can be sited remotely (in which case it is called an RSS). In both cases calls within the SSS can be switched directly or at the DLE. Thus all calls are switched in the same way whether or not the MDF is attached to the DLE or an RSS.
Where a call is being made to a subscriber on a different DLE from the caller, it needs to travel over some kind of link between DLEs. In theory all DLEs could be connected to one another, but this would be wasteful of links (there would be over a quarter of a million such links). Alternatively calls could be relayed from DLE to DLE (as was done in the days of manual operators), but this would take up a lot of the processing power of the DLEs.
Instead there is a second network of switches, known as the CORE NETWORK. This network is made up of four kinds of switch:
• 76 Digital Main Switching Units (DMSUs). These are the original long-distance exchanges; they are System X switches with the capacity to handle 30,000 simultaneous calls each. Every DMSU is connected to all the other DMSUs; obviously this involves a much smaller number of links than would be required to link together the DLEs.
• 16 Wide Area Tandems (WATs). A WAT is used to connect calls within a region, rather than longer-distance calls. While WATs are connected to other switches in the network, this links are less important. In particular, a WAT is not connected to all the DMSUs or even all the other WATs.
• 13 Digital Junction Switching Units (DJSUs). These are additional switches in the London area that allow traffic between London DLEs to be routed without taking up the capacity of the DMSUs and WATs.
• 63 Next Generation Switches (NGSs). These are newer switches that are gradually replacing the DMSUs and WATs, many of which have now reached the limits of their capacity. An NGS is a modified AXE10 capable of handling either 60,000 or 120,000 simultaneous calls (depending on configuration).
As well as all of these, there are also a certain number of inter-DLE links, particularly between DLEs in the same physical building.
Of course, BT is not the only telephone company in the UK. There are a range of Other Licensed Operators (OLOs) who also carry telephone calls. In order that calls can be transferred from BT to the OLO or vice versa, it is obviously necessary for the two networks to be connected. This is done at Points of Interconnection (POIs). A POI can be at any Core switch or, in some circumstances, at a DLE.
An OLO is not required to connect to every core switch. Instead, BT will route calls through the network to a convenient POI. While payment arrangements vary, the most common situation is that BT charges the OLO a transit fee for the call. This fee depends on the length of the call, the time of day, and the amount of the BT network that the call uses. For this purpose the most efficient path through the network is selected and the call is divided into one of a few classes:
• connected at the appropriate DLE (inter-DLE links are not used);
• single core;
• double core, further divided into "short", "medium", or "long"; every possible pair of core that could process the call is examined, and the shortest distance is used to determine the classification.
The PSTN is not a fully meshed network with every operator connected to every other - that would be both impractical and inefficient. Therefore calls may be routed through intermediate operator networks before they reach their final destination. One of the major problems in PSTN routing is determining how to route this call in the most cost effective and timely manner.
Each time a call is placed for routing, the desination number (also known as the called party) is entered by the calling party into their terminal. The destination number generally has two parts, a prefix which generally identifies the geographical location of the destination telephone, and a number unique within that prefix that determines the specific destination end. Sometimes if the call is between two customers in the same local area (that is, both customers are on the same telephone exchange, then the prefix may be omitted.
When a call is received by an exchange, there are two treatments that may be applied:
• Either the destination customer is directly connected to that exchange, in which case the call is placed down that connection and the destination customer’s phone rings.
• Or the call must be placed to one of the neighbouring exchanges through a connecting trunk for onward routing
Each exchange in the chain uses pre-calulated routing tables to determine which connected exchange the onward call should be routed to. There may be several alternative routes to any given destination, and the exchange can select dynamically between these in the event of link failure or congestion (see choice and proportional routing above)
The routing tables are generated centrally based on the known topology of the network, the numbering plan and analysis of traffic data. These are then downloaded to each exchange switch. Because of the hierarchical nature of the numbering plan, and its geographical basis, most calls can be routed based only on their prefix using these routing tables.
Some calls however cannot be routed on the basis of prefix alone, for example non-geographic numbers such as freephone numbers. In these cases they hit the intelligent network which supplies a translation which can be a PSTN number or a service address code.
In determining routing plans, special attention is paid for example to ensure that two routes do not mutually overflow to each other, otherwise congestion will cause a destination to be completely blocked.
According to BRAESS’S PARADOX the addition of a new, shorter, and lower cost route can lead to an increase overall congestion. The network planner must take this into account when designing routing paths.
One approach to routing involves the use of Dynamic Alternative Routing (DAR). DAR makes use of the distributed nature of a telecommunications network and its inherent randomness to dynamically determine optimal routing paths. This method generates a distributed, random, parallel computing platform that minimises congestion across the network, and is able to adapt to take changing traffic patterns and demands into account.
Braess's paradox, credited to the mathematician Dietrich Braess states that adding extra capacity to a network when the moving entities selfishly choose their route, can in some cases reduce overall performance. The paradox is stated as follows: "For each point of a road network, let there be given the number of cars starting from it, and the destination of the cars. Under these conditions one wishes to estimate the distribution of traffic flow. Whether one street is preferable to another depends not only on the quality of the road, but also on the density of the flow. If every driver takes the path that looks most favourable to him, the resultant running times need not be minimal. Furthermore, it is indicated by an example that an extension of the road network may cause a redistribution of the traffic that results in longer individual running times."
Finally, there are a number of special switches. The most important of these are the 26 Advanced Services Units (ASUs) which are Nortel DMS100 (or DMS10) switches used to implement things like FeatureNet. These are allocated various number ranges that don't necessarily match their physical location.
Each Core Network switch has a name, which usually includes the location of the switch. In addition, the DMSUs and WATs have names based on local themes, while the NGSs are named after minerals, precious metals, and precious and semi-precious stones, and the ASUs after mythological and astronomical features.
No comments:
Post a Comment